Notes on Quasi-Experimental Research & Ethics
Quasi-Experimental Research & Ethics — Study Notes
Purpose of today’s session
- Finish discussion of research methods by covering quasi-experimental designs
- Explore retrospective and prospective designs, developmental designs, and single-subject designs
- Briefly discuss ethics in research, including the Tuskegee syphilis study, the Belmont Report, and Institutional Review Boards (IRBs)
- Learning goals: define and give examples of quasi-experimental methods; differentiate retrospective vs prospective studies; name, describe, and differentiate Belmont Report principles
Quick course logistics (recap from transcript)
- Test scope: everything up to today’s class including today’s material
- Next test: approximately two lectures worth (e.g., lectures 1–5, depending on numbering)
- Instructor availability: email anytime; will respond when possible
Quick concept check (context from last class)
- Controlling for participant variables and demand characteristics
- Correct answer discussed: random assignment + double blind studies
- Key concepts:
- Participant variables: characteristics of a participant that can impact responses (e.g., fitness level, baseline stress)
- Demand characteristics: cues that influence participants to respond in a way they think the experimenter wants
- Double-blind design: both participants and experimenters are blind to the condition (which group a participant is in) to reduce bias
- Random assignment: randomly placing participants into experimental vs control conditions to avoid systematic differences
Quasi-experimental research: core idea
- Definition: a design that resembles an experimental design but lacks full random assignment
- Why it’s used: in health psychology and social science, it’s often impossible or unethical to manipulate certain variables
- Key feature: participants are grouped based on preexisting traits or conditions that cannot be randomly assigned
- Causal inference caveat: due to non-random assignment, stronger claims of causality are cautious; typically you infer associations rather than definitive causation
- Common approach: attempt to control for confounds through design or statistical methods, but cannot guarantee equivalence of groups
Core quasi-experimental designs in health psychology (three main categories taught)
- Retrospective and Prospective studies
- Developmental studies
- Single-subject designs
Retrospective vs Prospective designs
- Retrospective studies
- Look back after an outcome has occurred (e.g., disease, illness)
- Compare a group with the outcome to a similar group without the outcome
- Methods: examine medical records, work history, lived experiences to identify commonalities that might explain the outcome
- Real-world example discussed: nurses with noncancerous brain tumors clustered on a labor and delivery unit; retrospective look suggested a common occupational exposure
- Key limitations:
- Memory biases and inaccurate self-reports
- Missing or incomplete records
- Confounding factors not controllable after the fact
- Prospective studies
- Follow a cohort over time to see whether a development or outcome occurs
- Collect data on exposures and histories before the outcome occurs
- Example discussed: effect of smoking during pregnancy on birth weight; follow pregnant people over time and compare babies’ birth weights
- Key limitation: not randomized; other factors (e.g., SES, access to resources) may confound results
- Advantage: clearer temporal sequencing (exposure precedes outcome)
- Commonality: both are quasi-experimental because assignment to exposure groups is not randomized
Developmental quasi-experimental designs
- Focus on age-related factors; age is treated as a quasi-manipulated variable
- Cross-sectional designs
- Study different age groups at the same time
- Example: prevalence of a disease across age bands (e.g., 20–35, 36–50, 51–65)
- Limitation: cohort effects can confound interpretation
- Cohort effect: differences between age groups may reflect different historical exposures rather than aging per se
- Example discussion: younger cohorts may experience different environmental exposures than older cohorts, skewing observed relationships
- Practical notes: cost-efficient and fast, but less able to infer aging trajectories
- Longitudinal designs
- Follow the same group over time to observe development or change
- Advantage: can map developmental trajectories within individuals
- Disadvantage: expensive, lengthy, higher risk of attrition (participants drop out); resource-intensive
- Common practice: short-term longitudinal studies (e.g., 2–6 weeks) due to practicality
Single-subject designs
- Very small-N studies (often N = 1)
- Used when a participant has a rare condition or when individualized treatment effects are of interest
- Structure: use baseline (pre-treatment) as control; compare to outcomes after a treatment is introduced
- Distinction from case studies:
- Single-subject design is quasi-experimental with a within-subject comparison to baseline
- Case study is purely observational without a formal experimental manipulation or control condition
Example-driven clarification (smoking during pregnancy, birth weight, etc.)
- Quasi-experimental example: observing birth weights in babies of mothers who smoked vs. those who did not, without random assignment to smoking conditions
- Interpretation caution: observed associations may be due to other correlated factors (e.g., SES, nutrition, healthcare access)
Formulas and notation (basic design and causal thinking)
- Causal effect under random assignment (conceptual):
au = ext{E}[Y \,|\, X=1] - ext{E}[Y \,|\, X=0]
where Y = outcome (dependent variable), X = exposure/condition (1 = exposed, 0 = not exposed) - General causal model (conceptual):
Y = f(X) + \,\varepsilon
where \varepsilon captures unmeasured confounds and random error
- Causal effect under random assignment (conceptual):
Recap of key terms
- Participant variables: attributes of participants that can affect outcomes (e.g., fitness, prior experiences)
- Demand characteristics: cues that influence participants to respond in a way they think the study wants
- Double-blind: both participant and experimenter are unaware of which condition a participant is in
- Random assignment: randomly allocate participants to conditions to ensure equivalence across groups
- Retrospective vs Prospective: looking backward after an outcome vs following forward to observe future outcomes
- Cross-sectional vs Longitudinal: different age groups at one time vs same group across time
- Cohort effects: historical or environmental differences across age groups that confound age-related interpretations
Ethics in research: the Belmont Report, Tuskegee, and IRBs
- Tuskegee Syphilis Study (1932–1972)
- Conducted by Tuskegee University researchers with support from the U.S. government
- 400–600 impoverished Black men with syphilis and a comparison group; men were not informed of their condition and did not receive treatment even after penicillin became available
- Ethical violations included deception, lack of informed consent, withholding treatment, and exploitation of a vulnerable population
- Outcome: catalyzed reforms in human subjects research governance
- Belmont Report (1979) and its ethical principles
1) Respect for Persons
- Autonomy and informed consent; individuals should enter research voluntarily and can withdraw at any time
- Informed consent components include disclosure of risks, costs, benefits, and the right to withdraw
2) Beneficence - Maximize benefits while minimizing risks; weigh risks and benefits; protect participants from harm; consider privacy and confidentiality
- Examples of potential costs: time, mental/physical effort, possible discomfort; benefits can include knowledge, access to new therapies, or improvements in welfare
3) Justice - Fair and non-exploitative treatment; equitable selection of participants; ensure vulnerable populations are not exploited
- If a study targets a specific group, provide a theoretical justification for their inclusion; avoid excluding groups without rationale
- Informed consent and debriefing
- Informed consent: participants must be given sufficient information to decide; costs, benefits, potential harms, and withdrawal rights must be disclosed
- Debriefing: provided after participation, especially when deception was used; reveals true purpose of the study, offers support/resources if needed, and allows participants to withdraw their data if desired
- Institutional Review Boards (IRBs)
- Institutional committees (at universities, hospitals, etc.) that review all research involving human subjects before it begins
- Responsibilities: ensure participant rights and welfare; review consent forms, recruitment materials, debriefing scripts, and data handling procedures
- Requirement: typically mandated for federally funded research; there are analogous processes for animal research
- Practical ethics implications highlighted in class
- Deception requires thorough debriefing and justification; participants should be aware that deception occurred and whether data will be retained
- Respect for persons requires that coercive elements (e.g., excessive payment) are avoided; participation should be voluntary and informed
- Justice demands careful consideration of who is included in research and why; avoid exploiting vulnerable populations
Test-taking tips from today’s content
- Be able to distinguish quasi-experimental designs from true experiments: lack of random assignment is key
- Be able to differentiate retrospective vs prospective designs and identify their weaknesses and strengths
- Be able to name the Belmont Report’s three ethical principles and describe what each implies in practice
- Recognize the role of IRBs in approving and monitoring studies involving human participants
- Understand how participant variables and demand characteristics can bias results, and how double-blind and random assignment mitigate these issues
Real-world takeaways for exam preparation
- Recognize when a study is quasi-experimental based on lack of random assignment and ethical/practical constraints
- Distinguish cross-sectional versus longitudinal designs and identify cohort effects or attrition issues as needed
- Cite historical ethical breaches (e.g., Tuskegee) as motivators for modern ethical safeguards (Belmont Report, informed consent, debriefing, IRBs)
- Be comfortable discussing how ethical principles translate into everyday research design decisions (e.g., how to minimize harm, ensure fairness, and respect autonomy)
Quick glossary (to review before the exam)
- Quasi-experimental design: resembles an experiment but lacks random assignment
- Retrospective study: looks backward to identify causes after an outcome has occurred
- Prospective study: follows participants forward in time to observe outcomes
- Cross-sectional study: compares different age groups at one point in time
- Longitudinal study: follows the same group over time
- Single-subject design: N = 1 with within-subject comparisons
- Informed consent: voluntary, informed agreement to participate
- Debriefing: post-study explanation and support if deception or distress occurred
- IRB: Institutional Review Board, responsible for protecting human subjects in research
- Belmont Report: foundational document outlining respect for persons, beneficence, and justice
Final reminder from instructor
- For the upcoming test, you will be responsible for defining and giving examples of quasi-experimental methods, differentiating retrospective and prospective designs, and knowing the Belmont Report principles and IRB role
- You must be present to take the online Canva test
illustrative example recap (as reminders)
- Example: employment-related exposure studies after a cluster of illness outcomes emerged
- Example: prospective birth-control and cancer risk cohort study
- Example: real-world case where a deception-heavy study was debriefed afterward to mitigate potential harm
Equations to memorize (optional quick reference)
- Causal effect under random assignment: au = ext{E}[Y \,|\, X=1] - ext{E}[Y \,|\, X=0]
- General observation model: Y = f(X) + \varepsilon
End note
- The material integrates design choices (randomization, blinding, and control) with ethical safeguards (informed consent, debriefing, justice). Practice identifying the design type from a study description and articulating which Belmont principles apply in each case.